Research on Sentiment Analysis Based on Ensemble Learning

Abstract

The purpose of this paper is to use the ensemble learning method to analyze the sentiment of tourist attraction reviews, so as to improve the reference value of tourists in travel decision-making. With the popularity of online reviews, how to extract emotional information efficiently and accurately has become an important topic. In this study, random forest, gradient boosted tree (GBDT) and support vector machine (SVM) were used as the base learners, and the ensemble strategy of weighted voting was used to construct the model. Optimize parameters with cross-validation and grid search, and comprehensively evaluate model performance with metrics such as AUC, Log Loss, MCC, and average accuracy. Experimental results show that the classification accuracy and robustness of the ensemble model are better than those of the single model, with an AUC of 0.92 and an MCC of 0.78. The balance between precision and recall was demonstrated through performance surface plot analysis based on different thresholds, and the optimal threshold range was determined. The model can provide reliable sentiment insights for tourism platform users, help scenic spot managers identify user needs, optimize services, and have scalability in other review analysis fields.